Why sōns but absēns?

Question

The present participle of esse was (at one point) sōns, presumably from *h₁sonts. However, when a prefix is attached, it becomes -sēns, as in absēns and praesēns.

I'd always figured this was a relic of vowel reduction in non-initial syllables, like why we see facere but interficere. But Asteroides disagrees:

I don't think vowel reduction would actually result in sons/sont- changing to sens/sent- in a non-initial syllable. The usual reduction outcome of -o- in a closed medial or final syllable is -u-, as in the third-declension present 3rd person plural endings -unt, -untur or the old gerundive ending -undus. Sont- vs. -sent- look more like ablaut variants (I don't know the reason for that, but it could be as simple as analogy with other participles).

So: why do we see sōns next to -sēns? Is this how vowel reduction works before -ns (which we know lengthens vowels, and I can't think of any other instances of vowel reduction before it)? Or does ablaut or analogy factor in somehow?

Answer 1

Ruppel 2013 seems to offer the kind of explanation that I was thinking of, that absens and praesens were simply adapted to make their endings fit better with the usual morphology of Latin third-conjugation present participles:

The form we do find, sons, is used to mean 'guilty' ('he who (really) was it, he who is the real author of the thing in question'). In absens and praesens, the root vowel has been adapted to fit the vocalism of other regular present active participles. This is why I am referring to asterisked *sens in this section.

(page 121, Absolute Constructions in Early Indo-European, Antonia Ruppel, 2013)

A relevant question is obviously how other regular present active participles developed their vocalism, and what the vocalism was in Proto-Indo-European. Sihler 1995 says the following:

556. STEM ABLAUT. The full grade of the suffix *-ont- alternated with the zero grade *-n̥t- in the usual hysterokinetic way (full grade in nom./acc., zero grade in most other cases, 272.3, 273); the nom.sg. presents a special problem, about which more below (558). The ablaut of the suffix was the same for thematic and athematic types, that is, a complex like *bheront- 'carrying' is to be segmented *bher-ont-,¹ in contradistinction to the segmentation of the similar-looking thematic 3pl. *bher-o-nti. This indicates once upon a time the affix was in fact originally added to roots rather than stems, just like the verbal nouns. [...]

559. I. In ITALIC, uniquely, it was the zero grade of the stem that was generalized in the participles [...] and the resulting stems ferent-, legent- ousted all o-grade forms apart from the isolated sons, and eunt- (below). [...] Traces of the original o-grade can be found in voluntāt- 'will, wish', which enshrines *welont-; cf. volēns.

(pp. 613-616, New Comparative Grammar of Greek and Latin, Andrew L. Sihler)

If I'm reading the next quote correctly, it seems Jacob Wackernagel disputed that sons actually represents the participle of esse:

Not even the present participle in -ns is completely preserved in Latin. By no means every verb has it. Especially striking is the loss of *sens 'being'. Apart from the old sacral expression Di consentes (connected by the ancients with consentire ('agree with'), but rightly interpreted already generations ago by J. M. Gesner (1749) as meaning 'being with'), the participle of 'to be' survives only in absens and praesens (and in the latter with a shift of meaning). [...] The old suggestion that sons 'guilty' is the equivalent of Greek ὤν and means 'the one who is it' does not deserve to be pursued.

(pages 356-357, Participles in "Jacob Wackernagel, Lectures on Syntax: With Special Reference to Greek, Latin and Germanic", edited by David Langslow, 2009)

Vowel reduction

As for vowel reduction before -ns-, there appears to be an example in -scēns-, the supine stem for compounds of scando (ascendo, dēscendō, escendo). This could conceivably be an analogical formation from the vowel in the present stem -scend-, but we know from verbs like impingo, impāctum that analogy did not inevitably result in leveling of quality differences created by vowel reduction between the present and supine stems of verbs, so -scēns- seems to provide some support for ordering the rule of non-initial-syllable vowel reduction before the rule of pre-NS lengthening. Weiss 2009 cites the example of ascendere and gives an etymology for anhēlō that also requires this rule ordering: "*an-an-slō 'I pant' > *an-enslō > an(h)ēlō" (page 177, Outline of the Historical and Comparative Grammar of Latin, Michael Weiss).

But both of these examples show the outcome of the vowel -a-, which regularly yields -e- when reduced in closed syllables. As I mentioned, -o- in closed syllables regularly yields -u- when reduced. Although I know of no certain examples of this outcome where the -o- precedes /ns/, it hardly seems simpler to suppose a special phonetic development in this case than to explain the -ens in the nominative singular of absens, praesens and other third-conjugation participles by means of analogy, which seems to have caused leveling of many inflectional paradigms that originally featured ablaut (compare pēs vs. Greek πούς and English foot).